End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 2.1977350115776064e-06,
-    "train_runtime": 278.2719,
+    "train_loss": 2.6065856218338014e-06,
+    "train_runtime": 261.2528,
     "train_samples": 160,
-    "train_samples_per_second": 1.15,
-    "train_steps_per_second": 0.072
+    "train_samples_per_second": 1.225,
+    "train_steps_per_second": 0.077
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b382ce9227291f9f97f1a3a3300990ebae5eac665cffd7c90b5335b3ae2b4811
+oid sha256:4e0a4b7e4a48b737071c4fbf382d2ce78a803d21a72c0402e9ca41dbd52300ae
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 2.1977350115776064e-06,
-    "train_runtime": 278.2719,
+    "train_loss": 2.6065856218338014e-06,
+    "train_runtime": 261.2528,
     "train_samples": 160,
-    "train_samples_per_second": 1.15,
-    "train_steps_per_second": 0.072
+    "train_samples_per_second": 1.225,
+    "train_steps_per_second": 0.077
 }

trainer_state.json

@@ -10,203 +10,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 421.28125,
+      "completion_length": 455.09375,
       "epoch": 0.1,
-      "grad_norm": 5.338644504547119,
+      "grad_norm": 4.092990398406982,
       "kl": 0.0,
       "learning_rate": 4.965903258506806e-07,
       "loss": -0.0,
-      "reward": 2.437102566473186,
-      "reward_std": 2.4372674338519573,
-      "rewards/concensus_correctness_reward_func": 1.3522499799728394,
-      "rewards/consensus_reward_func": 0.0625,
+      "reward": 1.5772938383743167,
+      "reward_std": 1.2715626050485298,
+      "rewards/concensus_correctness_reward_func": 0.3084375001490116,
+      "rewards/consensus_reward_func": 0.3125,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.4409462600015104,
+      "rewards/question_recreation_reward_func": 0.47582510093343444,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.3314062627032399,
+      "rewards/xmlcount_reward_func": 0.23053125850856304,
       "step": 2
     },
     {
-      "completion_length": 473.1875,
+      "completion_length": 393.1875,
       "epoch": 0.2,
-      "grad_norm": 3.9548487663269043,
-      "kl": 0.0009631554330553627,
+      "grad_norm": 4.268237113952637,
+      "kl": 0.0010177667390962597,
       "learning_rate": 4.698684378016222e-07,
       "loss": 0.0,
-      "reward": 1.2341377083212137,
-      "reward_std": 0.6296644267858937,
-      "rewards/concensus_correctness_reward_func": 0.04143750108778477,
+      "reward": 1.5314793214201927,
+      "reward_std": 1.1808095844462514,
+      "rewards/concensus_correctness_reward_func": 0.16712500527501106,
       "rewards/consensus_reward_func": 0.125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.48288769845385104,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.43022931227460504,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.5223125051707029,
+      "rewards/xmlcount_reward_func": 0.6216250043362379,
       "step": 4
     },
     {
-      "completion_length": 380.125,
+      "completion_length": 489.53125,
       "epoch": 0.3,
-      "grad_norm": 3.4796371459960938,
-      "kl": 0.0015082741301739588,
+      "grad_norm": 7.274165630340576,
+      "kl": 0.0015757040673634037,
       "learning_rate": 4.193203929064353e-07,
       "loss": 0.0,
-      "reward": 2.055873781442642,
-      "reward_std": 1.7184474199893884,
-      "rewards/concensus_correctness_reward_func": 0.6996249984949827,
-      "rewards/consensus_reward_func": 0.125,
+      "reward": 3.2696286998689175,
+      "reward_std": 3.0783104081638157,
+      "rewards/concensus_correctness_reward_func": 1.4845624985173345,
+      "rewards/consensus_reward_func": 0.375,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.622748764231801,
+      "rewards/final_correctness_reward_func": 0.375,
+      "rewards/question_recreation_reward_func": 0.593472408130765,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.4835000205785036,
+      "rewards/xmlcount_reward_func": 0.44159374572336674,
       "step": 6
     },
     {
-      "completion_length": 390.625,
+      "completion_length": 391.28125,
       "epoch": 0.4,
-      "grad_norm": 5.360124111175537,
-      "kl": 0.001782346451363992,
+      "grad_norm": 4.208556175231934,
+      "kl": 0.001971208315808326,
       "learning_rate": 3.5042385616324236e-07,
       "loss": 0.0,
-      "reward": 1.2906564474105835,
-      "reward_std": 0.8444364763563499,
-      "rewards/concensus_correctness_reward_func": 0.10774999810382724,
+      "reward": 2.182107772678137,
+      "reward_std": 1.9605439552105963,
+      "rewards/concensus_correctness_reward_func": 0.773624999448657,
       "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.5392814450897276,
+      "rewards/question_recreation_reward_func": 0.4976702528074384,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.39362500002607703,
+      "rewards/xmlcount_reward_func": 0.6608125120401382,
       "step": 8
     },
     {
-      "completion_length": 432.5,
+      "completion_length": 354.875,
       "epoch": 0.5,
-      "grad_norm": 3.474890947341919,
-      "kl": 0.0018483960739104077,
+      "grad_norm": 4.434547424316406,
+      "kl": 0.003326518097310327,
       "learning_rate": 2.706448363680831e-07,
       "loss": 0.0,
-      "reward": 3.1597281731665134,
-      "reward_std": 2.401989583391696,
-      "rewards/concensus_correctness_reward_func": 1.4703124756924808,
-      "rewards/consensus_reward_func": 0.3125,
+      "reward": 2.21128673478961,
+      "reward_std": 1.2140152475330979,
+      "rewards/concensus_correctness_reward_func": 0.2775624953210354,
+      "rewards/consensus_reward_func": 0.5625,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3125,
-      "rewards/question_recreation_reward_func": 0.5964469783939421,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.5232867355225608,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.46796873956918716,
+      "rewards/xmlcount_reward_func": 0.7229375082533807,
       "step": 10
     },
     {
-      "completion_length": 360.34375,
+      "completion_length": 361.8125,
       "epoch": 0.6,
-      "grad_norm": 4.531510353088379,
-      "kl": 0.0027000612899428234,
+      "grad_norm": 4.211982250213623,
+      "kl": 0.003115422310656868,
       "learning_rate": 1.886286282148002e-07,
       "loss": 0.0,
-      "reward": 1.2967922370880842,
-      "reward_std": 0.518508761189878,
-      "rewards/concensus_correctness_reward_func": 0.03818750008940697,
-      "rewards/consensus_reward_func": 0.1875,
+      "reward": 1.6850453205406666,
+      "reward_std": 0.9321421514177928,
+      "rewards/concensus_correctness_reward_func": 0.13374999910593033,
+      "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.52576099766884,
+      "rewards/question_recreation_reward_func": 0.678232790902257,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.5453437687829137,
+      "rewards/xmlcount_reward_func": 0.6230625174939632,
       "step": 12
     },
     {
-      "completion_length": 361.40625,
+      "completion_length": 376.125,
       "epoch": 0.7,
-      "grad_norm": 4.74622106552124,
-      "kl": 0.0034341645368840545,
+      "grad_norm": 5.003601551055908,
+      "kl": 0.0039603932164027356,
       "learning_rate": 1.1326296046939333e-07,
       "loss": 0.0,
-      "reward": 2.393332852050662,
-      "reward_std": 2.1531090461649,
-      "rewards/concensus_correctness_reward_func": 0.8703750036656857,
-      "rewards/consensus_reward_func": 0.375,
+      "reward": 1.172058342024684,
+      "reward_std": 1.0172426409553736,
+      "rewards/concensus_correctness_reward_func": 0.08587499894201756,
+      "rewards/consensus_reward_func": 0.125,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.5497078304179013,
+      "rewards/question_recreation_reward_func": 0.36012083862442523,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.4732500038226135,
+      "rewards/xmlcount_reward_func": 0.47606251761317253,
       "step": 14
     },
     {
-      "completion_length": 335.96875,
+      "completion_length": 385.25,
       "epoch": 0.8,
-      "grad_norm": 5.897042274475098,
-      "kl": 0.003430365279200487,
+      "grad_norm": 4.454174995422363,
+      "kl": 0.003436649523791857,
       "learning_rate": 5.271487265090163e-08,
       "loss": 0.0,
-      "reward": 1.8474113084375858,
-      "reward_std": 1.2619444355368614,
-      "rewards/concensus_correctness_reward_func": 0.17718750052154064,
-      "rewards/consensus_reward_func": 0.25,
+      "reward": 4.223175309598446,
+      "reward_std": 3.490735905012116,
+      "rewards/concensus_correctness_reward_func": 2.179250003769994,
+      "rewards/consensus_reward_func": 0.4375,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3125,
-      "rewards/question_recreation_reward_func": 0.5919113333220594,
+      "rewards/final_correctness_reward_func": 0.4375,
+      "rewards/question_recreation_reward_func": 0.6258939877152443,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.5158125096932054,
+      "rewards/xmlcount_reward_func": 0.5430312510579824,
       "step": 16
     },
     {
-      "completion_length": 432.71875,
+      "completion_length": 380.46875,
       "epoch": 0.9,
-      "grad_norm": 4.743912220001221,
-      "kl": 0.003188079885148909,
+      "grad_norm": 4.106444835662842,
+      "kl": 0.0034525456139817834,
       "learning_rate": 1.3545689574841341e-08,
       "loss": 0.0,
-      "reward": 1.8755912049673498,
-      "reward_std": 1.678645506966859,
-      "rewards/concensus_correctness_reward_func": 0.7280000001192093,
-      "rewards/consensus_reward_func": 0.125,
+      "reward": 1.6361163649708033,
+      "reward_std": 1.052419565152377,
+      "rewards/concensus_correctness_reward_func": 0.28506250120699406,
+      "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.42705990676768124,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.561085123103112,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.47053124965168536,
+      "rewards/xmlcount_reward_func": 0.5399687550961971,
       "step": 18
     },
     {
-      "completion_length": 368.375,
+      "completion_length": 393.1875,
       "epoch": 1.0,
-      "grad_norm": 5.017685890197754,
-      "kl": 0.0036494637379291817,
+      "grad_norm": 3.7455756664276123,
+      "kl": 0.004411991249071434,
       "learning_rate": 0.0,
       "loss": 0.0,
-      "reward": 2.507398460060358,
-      "reward_std": 1.567327823897358,
-      "rewards/concensus_correctness_reward_func": 0.6983124995604157,
+      "reward": 2.0187183283269405,
+      "reward_std": 0.8147448373201769,
+      "rewards/concensus_correctness_reward_func": 0.0521249994635582,
       "rewards/consensus_reward_func": 0.1875,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.375,
-      "rewards/question_recreation_reward_func": 0.6738984161056578,
+      "rewards/final_correctness_reward_func": 0.5,
+      "rewards/question_recreation_reward_func": 0.6972183138132095,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.5726875076070428,
+      "rewards/xmlcount_reward_func": 0.5818750094622374,
       "step": 20
     },
     {
       "epoch": 1.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 2.1977350115776064e-06,
-      "train_runtime": 278.2719,
-      "train_samples_per_second": 1.15,
-      "train_steps_per_second": 0.072
+      "train_loss": 2.6065856218338014e-06,
+      "train_runtime": 261.2528,
+      "train_samples_per_second": 1.225,
+      "train_steps_per_second": 0.077
     }
   ],
   "logging_steps": 2,